Non-Stochastic Hypothesis Testing with Application to Privacy Against Hypothesis-Testing Adversary

TL;DR

This paper introduces a non-stochastic hypothesis testing framework to quantify and enhance privacy against adversaries, proposing new bounds, privacy measures, and reporting policies evaluated on real demographic data.

Contribution

It develops a non-stochastic hypothesis testing theory, defines a privacy measure, and constructs privacy-preserving reporting policies with utility guarantees.

Findings

Established a fundamental bound on non-stochastic hypothesis tests

Proposed a privacy measure based on this bound

Demonstrated privacy-utility trade-offs on real demographic data

Abstract

In this paper, we consider privacy against hypothesis testing adversaries within a non-stochastic framework. We develop a theory of non-stochastic hypothesis testing by borrowing the notion of uncertain variables from non-stochastic information theory. We define tests as binary-valued mappings on uncertain variables and prove a fundamental bound on the best performance of tests in non-stochastic hypothesis testing. We use this bound to develop a measure of privacy. We then construct reporting policies with prescribed privacy and utility guarantees. The utility of a reporting policy is measured by the distance between the reported and original values. We illustrate the effects of using such privacy-preserving reporting polices on a publicly-available practical dataset of preferences and demographics of young individuals, aged between 15-30, with Slovakian nationality.